Table 2. Performance of classification of CXRs dataset, DB using various feature extraction methods.
Feature* | Accuracy | Sensitivity | Specificity | Precision | F-score | GMean | AUC | CI# | SE# |
Gab1 | 0.920 | 0.920 | 0.920 | 0.920 | 0.920 | 0.920 | 0.936 | 0.829–0.986 | 0.0377 |
Gist1 | 0.860 | 0.880 | 0.840 | 0.846 | 0.863 | 0.860 | 0.893 | 0.773–0.962 | 0.0520 |
HOG1 | 0.860 | 0.800 | 0.920 | 0.909 | 0.851 | 0.858 | 0.909 | 0.793–0.972 | 0.0434 |
PHOG1 | 0.900 | 0.920 | 0.880 | 0.885 | 0.902 | 0.900 | 0.918 | 0.806–0.977 | 0.0404 |
*Features extracted from: 1 indicates whole CXR, 2 indicates manually segmented CXR and 3 indicates automatic segmented CXR.
CI refers to confidence interval at 95% at P-value <0.0001 whereas SE refers to standard error for AUC.